Google Cloud Introduces C4N General Availability with Industry-Leading Network and Storage Performance

As enterprise applications become increasingly data intensive, compute power alone is no longer enough. Modern databases, AI inference, cybersecurity appliances, analytics platforms, and telecommunications workloads demand exceptionally high networking throughput and storage performance. Traditional virtual machines often struggle to keep pace, forcing organizations to overprovision compute resources simply to meet I/O requirements.

To address this challenge, Google Cloud has officially announced the general availability (GA) of C4N, its first network- and block-storage-optimized Compute Engine machine family. Initially unveiled during Google Cloud Next '26, C4N is purpose-built for workloads where networking and storage performance are critical.

Powered by Google's Titanium offload architecture and 5th Generation Intel Xeon Scalable (Emerald Rapids) processors, C4N delivers up to 400 Gbps of networking bandwidth, 95 million packets per second (MPPS), and 1 million Hyperdisk Extreme IOPS, making it Google's highest-performing x86 virtual machine for network-intensive applications.

Why Google Built C4N

Enterprise infrastructure has evolved dramatically over the past decade.

Organizations now operate:

• Large-scale distributed databases

• AI and machine learning inference platforms

• High-performance analytics engines

• Software-defined networking

• 5G core infrastructure

• Security appliances

• Real-time streaming applications

These workloads continuously move enormous amounts of data between servers, storage systems, and users.

In many cases, networking performance, not CPU utilisation, becomes the primary bottleneck.

C4N addresses this problem by separating networking and storage processing from application compute through Google's custom Titanium offload architecture, allowing CPUs to focus on application execution. At the same time, dedicated hardware manages network and storage operations.

The result is dramatically higher throughput, lower latency, and improved compute efficiency.

Performance That Redefines Cloud Networking

Google positions C4N as its highest-performing x86 Compute Engine instance for networking workloads.

Key capabilities include:

• Up to 400 Gbps VM-to-VM network bandwidth

• Up to 200 Gbps internet egress bandwidth

• Up to 95 million packets per second

• Up to 50 Gbps single-flow bandwidth

• Nearly 33% more network bandwidth per vCPU compared to similar Intel-based hyperscaler offerings

• Up to 224% faster packet processing performance

Unlike many competing cloud offerings that require premium networking add-ons, C4N delivers these capabilities as standard.

This makes the platform particularly attractive for organizations running bandwidth-intensive applications without additional licensing complexity.

Built for Modern Enterprise Workloads

C4N is optimized for applications that require predictable high-throughput networking and storage performance.

Ideal workloads include:

Network and Security Appliances

• Next-generation firewalls

• Virtual routers

• Load balancers

• DDoS mitigation platforms

• VPN gateways

Data Platforms

• Distributed databases

• High-performance file systems

• Data lakes

• Real-time analytics

• Big data processing

Telecommunications

Cloud-native 5G Core functions, including User Plane Function (UPF), benefit from C4N's high packet processing capabilities and deterministic networking.

Artificial Intelligence

CPU-based AI inference, machine learning pipelines, and distributed AI infrastructure can leverage higher network throughput while efficiently accessing massive datasets.

Hyperdisk Unlocks Exceptional Storage Performance

Networking alone is only part of the equation.

Storage-intensive applications often require millions of I/O operations per second while maintaining low latency.

When paired with Hyperdisk Extreme, C4N delivers Compute Engine's highest block storage performance.

Highlights include:

• Up to 25 GiB/s storage throughput

• Up to 1 million IOPS

• Nearly 33% higher storage bandwidth per vCPU

• Approximately 39% more IOPS compared with similar Intel-based cloud offerings

Organizations can independently scale storage performance without resizing compute resources, significantly improving infrastructure efficiency.

Google also supports the full Hyperdisk portfolio on C4N:

• Hyperdisk Balanced

• Hyperdisk Balanced High Availability

• Hyperdisk Extreme

• Hyperdisk Throughput

• Hyperdisk ML

This flexibility enables organizations to optimize performance for different workload requirements.

Faster Networking Across Every Deployment

C4N introduces several networking enhancements beyond raw bandwidth.

Superior VM-to-VM Performance

Internal networking scales up to 400 Gbps, nearly four times the bandwidth per vCPU compared to standard C4 instances.

Faster Internet Connectivity

Internet egress bandwidth increases up to 200 Gbps, while internet packet processing performance improves nearly 32 times, reaching 48 MPPS.

Better Performance for Smaller Instances

Even compact C4N machine types with only 2 to 16 vCPUs receive dedicated networking performance ranging from 25 Gbps to 50 Gbps, reducing the need to purchase oversized virtual machines solely for networking capacity.

Enhanced gVNIC Networking

Google has also increased the default number of transmit and receive queues available through gVNIC interfaces, improving parallel networking performance without manual tuning.

Improved Cloud Storage Transfers

C4N doubles bandwidth when transferring data to and from Google Cloud Storage, benefiting:

• AI training

• Machine learning inference

• Data analytics

• Backup operations

• Large-scale data ingestion

Real-World Application Performance

Google shared benchmark improvements demonstrating C4N's capabilities across common enterprise workloads.

Examples include:

Web Serving

Up to 1.5× more NGINX requests per second compared with standard C4 virtual machines.

Database Performance

Up to 45% higher MySQL queries per second when datasets primarily reside on disk.

These improvements allow organizations to handle greater workloads while reducing infrastructure costs.

Customer Success Stories

Several industry leaders are already deploying C4N across demanding enterprise environments.

Ericsson

Ericsson uses C4N to power its cloud-native 5G Core-as-a-Service, achieving 1 Tbps throughput while maintaining carrier-grade reliability for telecommunications workloads.

Teradata

Teradata integrates C4N into its Autonomous Knowledge Platform to improve price-performance for AI, analytics, and mission-critical enterprise databases.

NetApp

NetApp is expanding support for C4N to accelerate data-in-place AI and analytics using Google Cloud NetApp Volumes.

Sycomp

In internal testing, Sycomp achieved:

• 58.5 GiB/s read throughput using three storage servers

• 195 GiB/s read and write throughput using ten storage servers

These results reached approximately 97% of theoretical maximum storage performance.

ClipperDB Technologies

Running industry-standard TPC-DS analytics benchmarks, ClipperDB reported:

• More than 3× lower cost per query

• Up to 11× faster Spark analytics

while maintaining full Apache Spark compatibility.

Flexible Configurations for Every Enterprise

The C4N machine family offers extensive deployment flexibility.

Available configurations include:

• 2 to 192 vCPUs

• Up to 1.5 TB DDR5 memory

• High CPU

• Standard

• High Memory

• Local SSD variants

Future enhancements will include:

• Up to 12 TiB Titanium Local SSD

• Native bare-metal C4N servers

• Expanded service support beyond Compute Engine and Google Kubernetes Engine (GKE)

These options allow organizations to tailor infrastructure for everything from lightweight networking appliances to memory-intensive analytics clusters.

Lower Total Cost of Ownership

One of C4N's biggest advantages is eliminating unnecessary overprovisioning.

Traditionally, organizations needing higher networking throughput often had to purchase significantly larger virtual machines simply to obtain additional bandwidth.

C4N separates compute performance from networking and storage scaling, enabling organizations to provision resources more precisely.

This results in:

• Better infrastructure utilization

• Lower operational costs

• Reduced cloud spending

• Higher workload density

• Improved performance predictability

For enterprises running large-scale production environments, these efficiencies can substantially reduce the total cost of ownership over time.

Final Thoughts

The general availability of Google Cloud C4N marks an important milestone for infrastructure designed around data movement rather than raw compute alone.

By combining Titanium offload architecture, 5th Generation Intel Xeon processors, 400 Gbps networking, and Hyperdisk Extreme, Google delivers one of the industry's most capable x86 cloud platforms for networking, storage, analytics, AI inference, and telecommunications workloads.

As enterprise applications continue demanding higher throughput with predictable performance, C4N provides organizations with a scalable foundation that improves efficiency, reduces infrastructure bottlenecks, and lowers total cost of ownership across modern cloud deployments.